Maskless lithography system and lithography method thereof

Abstract

The present application relates to the field of intelligent detection technology, and specifically discloses a maskless lithography system and a lithography method thereof, which uses ultraviolet light as the initial light source, exposes the image data of the microelectronic pattern to the PCB board for etching through FPGA programming and DMD, and uses machine vision technology based on deep learning to measure the global semantic consistency of the image data of the etched solidified pattern and the microelectronic pattern, so as to intelligently detect whether the solidified pattern meets the design requirements. In this way, the lithography accuracy and stability can be effectively improved, the quality of the lithography products can be guaranteed, and the demand for high-precision, high-efficiency, and low-cost manufacturing in the field of microelectronics can be met.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection technology, and more specifically, to a maskless lithography system and a lithography method thereof. Background Art

[0002] Photolithography is a key technology in the field of microelectronics manufacturing. It uses light to illuminate specially treated photosensitive materials and transfer the fine structure of microelectronic circuits to semiconductor wafers or other substrates through the pattern on the mask. With the rapid development of science and technology, the requirements for the manufacturing accuracy and efficiency of PCB boards in the field of microelectronics are increasing. As the core process in microelectronics manufacturing, the accuracy and efficiency of photolithography directly affect the performance and market competitiveness of microelectronic products.

[0003] In traditional mask lithography, a physical mask is needed to limit the penetration range of light to achieve pattern transfer. However, the mask production process is complex and costly, with problems such as long manufacturing cycle, high cost, and poor flexibility, which cannot meet the growing demand for personalization and rapid iteration in the field of microelectronics.

[0004] In recent years, maskless lithography has gradually become a research hotspot due to its advantages such as high efficiency, flexibility and low cost. Maskless lithography generates digital mask patterns through computers, and uses precise optical systems and control algorithms to achieve precise processing of micron- and even nanoscale structures, providing new solutions for the field of micro-nano manufacturing. However, some existing maskless lithography technologies face modulator resolution limitations, which affect lithography accuracy and stability. Therefore, an optimized maskless lithography system and method are expected. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of the present application provides a maskless lithography system and a lithography method thereof, which uses ultraviolet light as the initial light source, exposes the image data of the microelectronic pattern to the PCB board for etching through FPGA programming and DMD, and uses machine vision technology based on deep learning to measure the global semantic consistency of the image data of the etched solidified pattern and the microelectronic pattern, so as to intelligently detect whether the solidified pattern meets the design requirements. In this way, the lithography accuracy and stability can be effectively improved, the quality of lithography products can be guaranteed, and the demand for high-precision, high-efficiency and low-cost manufacturing in the field of microelectronics can be met.

[0006] Accordingly, according to one aspect of the present application, a maskless lithography method is provided, which includes: providing image data of a microelectronic pattern; inputting the image data of the microelectronic pattern into a digital micromirror device, wherein the digital micromirror device includes a plurality of rotatable mirrors; preparing an ultraviolet light source; programming and controlling the digital micromirror device through a field programmable gate array so that the plurality of rotatable mirrors of the digital micromirror device can reflect the ultraviolet light generated by the ultraviolet light source according to instructions; modulating the ultraviolet light using a coated lens so that the modulated ultraviolet light is irradiated onto a PCB board coated with a photosensitive material to obtain an exposed PCB board; developing and etching the exposed PCB board to obtain a cured pattern; and inspecting the cured pattern based on machine vision to ensure that it meets the design requirements.

[0007] In the above-mentioned maskless lithography method, the cured pattern is inspected based on machine vision to ensure that it meets the design requirements, including: obtaining a digital image of the cured pattern captured by a camera; performing local area multi-scale semantic analysis on the digital image of the cured pattern and the image data of the microelectronic pattern to obtain a set of local area semantic feature vectors of the cured pattern and a set of local area semantic feature vectors of the design pattern; performing global semantic consistency measurement on the set of local area semantic feature vectors of the cured pattern and the set of local area semantic feature vectors of the design pattern to obtain a global semantic consistency representation vector; and determining the detection result based on the global semantic consistency representation vector.

[0008] In the above-mentioned maskless lithography method, the digital image of the solidified pattern and the image data of the microelectronic pattern are respectively subjected to local area multi-scale semantic analysis to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern, including: segmenting the digital image of the solidified pattern and the image data of the microelectronic pattern in the same image block segmentation mode to obtain a set of local area images of the solidified pattern and a set of local area images of the design pattern; performing image feature extraction on the set of local area images of the solidified pattern and the set of local area images of the design pattern to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern; passing the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern through a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern.

[0009] In the above-mentioned maskless lithography method, image feature extraction is performed on the set of local area images of the solidified pattern and the set of local area images of the design pattern respectively to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern, including: the set of local area images of the solidified pattern and the set of local area images of the design pattern are respectively passed through a feature extractor based on a void convolutional neural network model to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern.

[0010] In the above-mentioned maskless lithography method, the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern are respectively passed through a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern, including: respectively extracting the global semantic features and local semantic features of the local area feature maps of the solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern; fusing the global semantic feature vector of the local area of ​​the solidified pattern and the local semantic feature vector of the local area of ​​the solidified pattern to obtain the local area semantic feature vector of the solidified pattern.

[0011] In the above-mentioned maskless lithography method, the global semantic features and local semantic features of the local area feature map of the solidified pattern are respectively extracted to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern, including: upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; performing global mean pooling processing on the upsampled local area feature map of the solidified pattern to obtain an upsampled local area semantic channel feature vector of the solidified pattern; performing point convolution encoding on the upsampled local area semantic channel feature vector of the solidified pattern to obtain the global semantic feature vector of the local area of ​​the solidified pattern.

[0012] In the above-mentioned maskless lithography method, the global semantic features and local semantic features of the local area feature map of the solidified pattern are respectively extracted to obtain the global semantic feature vector of the local area of ​​the solidified pattern and the local semantic feature vector of the local area of ​​the solidified pattern, and also include: upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; performing point convolution encoding on the upsampled local area feature map of the solidified pattern to obtain a channel-modulated upsampled local area feature map of the solidified pattern; performing two-dimensional convolution encoding on the channel-modulated upsampled local area feature map of the solidified pattern to obtain the local semantic feature vector of the local area of ​​the solidified pattern.

[0013] In the above-mentioned maskless lithography method, a global semantic consistency measurement is performed on the set of local area semantic feature vectors of the solidified pattern and the set of local area semantic feature vectors of the design pattern to obtain a global semantic consistency representation vector, including: calculating the hash similarity between each group of corresponding local area semantic feature vectors of the solidified pattern and the local area semantic feature vectors of the design pattern in the set of local area semantic feature vectors of the solidified pattern and the set of local area semantic feature vectors of the design pattern to obtain the global semantic consistency representation vector composed of multiple hash similarities.

[0014] In the above-mentioned maskless lithography method, based on the global semantic consistency representation vector, the detection result is determined, including: passing the global semantic consistency representation vector through a classifier-based detector to obtain the detection result, and the detection result is used to indicate whether the solidified pattern meets the design requirements.

[0015] According to another aspect of the present application, a maskless lithography system is provided, which includes: a curing pattern digital image acquisition module, used to acquire image data of a microelectronic pattern, and a digital image of a curing pattern captured by a camera; a local area multi-scale semantic analysis module, used to perform local area multi-scale semantic analysis on the digital image of the curing pattern and the image data of the microelectronic pattern, respectively, to obtain a set of local area semantic feature vectors of the curing pattern and a set of local area semantic feature vectors of the design pattern; a global semantic consistency measurement module, used to perform global semantic consistency measurement on the set of local area semantic feature vectors of the curing pattern and the set of local area semantic feature vectors of the design pattern to obtain a global semantic consistency representation vector; and a curing pattern detection result generation module, used to determine a detection result based on the global semantic consistency representation vector.

[0016] Compared with the prior art, the maskless lithography system and lithography method provided by the present application use ultraviolet light as the initial light source, expose the image data of the microelectronic pattern to the PCB board for etching through FPGA programming and DMD, and use the machine vision technology based on deep learning to measure the global semantic consistency of the image data of the etched solidified pattern and the microelectronic pattern, so as to intelligently detect whether the solidified pattern meets the design requirements. In this way, the lithography accuracy and stability can be effectively improved, the quality of lithography products can be guaranteed, and the demand for high-precision, high-efficiency, and low-cost manufacturing in the field of microelectronics can be met. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 The present invention is a flow chart of inspecting the cured pattern based on machine vision in the maskless lithography method according to an embodiment of the present application to ensure that the cured pattern meets the design requirements.

[0019] Figure 2 It is a schematic diagram of an architecture for inspecting the solidified pattern based on machine vision in a maskless lithography method according to an embodiment of the present application to ensure that it meets the design requirements.

[0020] Figure 3 The present invention is a flowchart of performing local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern in the maskless lithography method according to an embodiment of the present application to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern.

[0021] Figure 4 The present invention is a flowchart of obtaining a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern in a maskless lithography method according to an embodiment of the present application, respectively, by passing the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern through a multi-scale semantic feature extractor.

[0022] Figure 5 4 is a block diagram of a maskless lithography system according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the embodiments of the present application will be described in more detail in conjunction with the accompanying drawings, and the above and other purposes, features and advantages of the present application will become more apparent. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0024] As mentioned in the above background technology, in recent years, maskless lithography technology generates digital mask graphics through computers, and uses precision optical systems and control algorithms to achieve precise processing of micron and even nanometer-level structures. With its high efficiency, flexibility and low cost, it has gradually become a research focus. However, some existing maskless lithography technologies are limited by the resolution of the modulator, which affects the lithography accuracy and stability. In view of the above technical problems, the technical concept of this application is to use ultraviolet light as the initial light source, expose the image data of the microelectronic pattern to the PCB board for etching through FPGA programming and DMD, and use the machine vision technology based on deep learning to measure the global semantic consistency of the image data of the cured pattern and the microelectronic pattern after etching, so as to intelligently detect whether the cured pattern meets the design requirements. In this way, it is possible to effectively improve the lithography accuracy and stability, ensure the quality of lithography products, and meet the needs of the microelectronics field for high-precision, high-efficiency and low-cost manufacturing.

[0025] Based on this, the present application proposes a maskless lithography method, which includes the steps of: providing image data of a microelectronic pattern; inputting the image data of the microelectronic pattern into a digital micromirror device, wherein the digital micromirror device includes a plurality of rotatable mirrors; preparing an ultraviolet light source; programming and controlling the digital micromirror device through a field programmable gate array so that the plurality of rotatable mirrors of the digital micromirror device can reflect the ultraviolet light generated by the ultraviolet light source according to instructions; modulating the ultraviolet light using a coated lens so that the modulated ultraviolet light is irradiated onto a PCB board coated with a photosensitive material to obtain an exposed PCB board; developing and etching the exposed PCB board to obtain a cured pattern; and inspecting the cured pattern based on machine vision to ensure that it meets the design requirements.

[0026] Specifically, a digital micromirror device (DMD) is an optical device based on MEMS (Micro-electromechan-System) technology, which can control light by controlling the rotation angle of a micromirror. Maskless lithography is a new type of micro-nano processing technology. Compared with traditional mask lithography, it can change the drawing pattern at any time as needed, without remaking the mask, and has the advantages of simple process, high processing efficiency, and low manufacturing cost. Combining digital micromirror devices with maskless lithography technology can produce a high-speed spatial light modulator. Specifically, in the technical solution of the present application, the rotation angle of the micromirror is controlled by FPGA programming, so that the grating structure is displaced, thereby realizing spatial modulation of light waves. Compared with traditional spatial light modulators, digital micromirror devices can achieve sub-pixel level precision control and microsecond response time, and can realize rapid preparation of micron-level optical structures, with ultra-high-speed modulation capability and extremely high precision.

[0027] In addition, after the photolithography is completed, in order to ensure that the quality of the cured pattern meets the design requirements, the cured pattern needs to be inspected. Traditional inspection methods are usually based on manual visual inspection or simple image processing methods. These methods have the problems of strong subjectivity, low efficiency, and insufficient precision. Therefore, in the technical solution of the present application, machine vision technology is further used to inspect the cured pattern. By comparing and analyzing the cured pattern and the microelectronic pattern, the similarity of the image semantic features between the two is excavated, so as to intelligently identify whether the cured pattern meets the design requirements. In this way, the detection efficiency and accuracy can be effectively improved, and the quality and stability of the photolithography products can be guaranteed.

[0028] Figure 1 The present invention is a flow chart of inspecting the cured pattern based on machine vision in the maskless lithography method according to an embodiment of the present application to ensure that the cured pattern meets the design requirements. Figure 2 FIG. 1 is a schematic diagram of a structure for inspecting the solidified pattern based on machine vision in a maskless lithography method according to an embodiment of the present application to ensure that the solidified pattern meets the design requirements. Figure 1 and Figure 2As shown, the maskless lithography method according to the embodiment of the present application includes the steps of: S110, acquiring a digital image of the solidified pattern captured by a camera; S120, performing local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern; S130, performing global semantic consistency measurement on the set of local area semantic feature vectors of the solidified pattern and the set of local area semantic feature vectors of the design pattern to obtain a global semantic consistency representation vector; S140, determining a detection result based on the global semantic consistency representation vector.

[0029] In the above-mentioned maskless lithography method, the step S110 is to obtain a digital image of the solidified pattern captured by a camera. It should be understood that the camera can quickly and in real time capture the image of the solidified pattern, so that the entire detection process can be automated without manual intervention, and the digital image captured by the camera is easy to store, transmit and calculate, providing a basis for implementing image processing and analysis algorithms. In addition, image acquisition by a camera is a non-contact detection method that does not cause any physical damage or contamination to the solidified pattern, thereby ensuring the integrity and reliability of the solidified pattern.

[0030] In the above-mentioned maskless lithography method, in step S120, the digital image of the solidified pattern and the image data of the microelectronic pattern are respectively subjected to local area multi-scale semantic analysis to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern. Figure 3 The present invention is a flowchart of performing local multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern in the maskless lithography method according to an embodiment of the present application to obtain a set of local semantic feature vectors of the solidified pattern and a set of local semantic feature vectors of the design pattern. Figure 3 As shown, the step S120 includes: S121, segmenting the digital image of the solidified pattern and the image data of the microelectronic pattern in the same image block segmentation mode to obtain a set of local area images of the solidified pattern and a set of local area images of the design pattern; S122, performing image feature extraction on the set of local area images of the solidified pattern and the set of local area images of the design pattern to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern; S123, passing the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern through a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern.

[0031] Specifically, in step S121, the digital image of the solidified pattern and the image data of the microelectronic pattern are segmented in the same image block segmentation mode to obtain a set of local area images of the solidified pattern and a set of local area images of the design pattern. It should be understood that, considering that in the comparative analysis process of the digital image of the solidified pattern and the image data of the microelectronic pattern, if the whole image is directly subjected to global comparative analysis, some differences in local details may be ignored, and the amount of calculation will be too large, affecting the processing speed and efficiency. Therefore, in the technical solution of the present application, the digital image of the solidified pattern and the image data of the microelectronic pattern are segmented in the same image block segmentation mode to divide the image into multiple local areas. By performing a separate comparative analysis on each local area image, the local detail differences between the solidified pattern and the design pattern can be more accurately captured, and the amount of calculation for a single image processing is reduced, thereby improving the response speed and real-time performance of the entire system. In addition, by adopting the same image block segmentation mode, it can be ensured that the local areas of the solidified pattern and the design pattern are consistent in position, size and shape, so that in the subsequent feature extraction and semantic feature comparison process, the two can be matched and compared more accurately.

[0032] Specifically, in step S122, image feature extraction is performed on the set of local area images of the solidified pattern and the set of local area images of the design pattern to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern. In a specific example of the present application, the processing method of performing image feature extraction on the set of local area images of the solidified pattern and the set of local area images of the design pattern is to pass the set of local area images of the solidified pattern and the set of local area images of the design pattern through a feature extractor based on a hole convolutional neural network model to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the design pattern. It should be known to those skilled in the art that the hole convolution (also called dilated convolution or atrous convolution) can effectively expand the receptive field of the convolution kernel and capture more spatial context information without increasing the number of parameters and computational cost by inserting additional zero values ​​in the convolution kernel. Here, the feature extractor based on the atrous convolutional neural network model utilizes the characteristics of atrous convolution. Through convolution and pooling operations, it can extract useful feature information from the local area image of the solidified pattern and the local area image of the designed pattern, and mine the image feature representations such as color, texture, shape, and structure of the image, thereby achieving effective representation of the image content and providing strong data support for subsequent image feature similarity measurement.

[0033] Specifically, in step S123, the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern are respectively passed through a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern. It should be understood that, although the feature representation of the image can be extracted through the processing of the hole convolutional neural network, it usually only contains the underlying information of the image and is difficult to be directly used to describe the high-level semantic content of the image. Therefore, in the technical solution of the present application, a multi-scale semantic feature extractor is further introduced to further process and abstract the set of local area feature maps of the solidified pattern and the set of local area feature maps of the design pattern. Specifically, the multi-scale semantic feature extractor extracts and fuses the global semantic features and local semantic features of the local area feature maps to achieve a semantic description of the image content. Among them, global semantics represents the overall meaning or concept of the image and provides overall background information, while local semantics represents the detailed meaning of a specific area or object and provides detail information. By fusing the global semantics and local semantics of an image, we can improve our deep semantic understanding of the image, thereby improving the accuracy of subsequent comparative analysis of image semantic features.

[0034] Figure 4 The present invention is a flowchart of obtaining a set of local region semantic feature vectors of the solidified pattern and a set of local region semantic feature vectors of the designed pattern by respectively passing the set of local region feature maps of the solidified pattern and the set of local region feature maps of the designed pattern through a multi-scale semantic feature extractor in a maskless lithography method according to an embodiment of the present application. Figure 4 As shown, the step S123 includes: S1231, respectively extracting the global semantic features and the local semantic features of the local area feature map of the solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern; S1232, fusing the global semantic feature vector of the local area of ​​the solidified pattern and the local semantic feature vector of the local area of ​​the solidified pattern to obtain the semantic feature vector of the local area of ​​the solidified pattern.

[0035] More specifically, the step S1231 includes: upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; performing global mean pooling processing on the upsampled local area feature map of the solidified pattern to obtain an upsampled local area semantic channel feature vector of the solidified pattern; performing point convolution encoding on the upsampled local area semantic channel feature vector of the solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern.

[0036] More specifically, the step S1231 also includes: upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; performing point convolution encoding on the upsampled local area feature map of the solidified pattern to obtain a channel-modulated upsampled local area feature map of the solidified pattern; performing two-dimensional convolution encoding on the channel-modulated upsampled local area feature map of the solidified pattern to obtain a local semantic feature vector of the local area of ​​the solidified pattern.

[0037] In the above-mentioned maskless lithography method, the step S130 performs global semantic consistency measurement on the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the design pattern to obtain a global semantic consistency representation vector. In a specific example of the present application, the processing method for performing global semantic consistency measurement on the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the design pattern is to calculate the hash similarity between each group of corresponding semantic feature vectors of the local area of ​​the solidified pattern and the semantic feature vectors of the local area of ​​the design pattern in the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the design pattern to obtain the global semantic consistency representation vector composed of multiple hash similarities. That is, in order to quantitatively evaluate the semantic consistency between each local area of ​​the solidified pattern and each local area of ​​the design pattern, the degree of semantic similarity between the two is measured by calculating the hash similarity between each group of corresponding semantic feature vectors of the local area of ​​the solidified pattern and the semantic feature vectors of the local area of ​​the design pattern. Specifically, the hash function can map each group of corresponding semantic feature vectors of the local area of ​​the solidified pattern and the semantic feature vectors of the local area of ​​the design pattern into the hash space to obtain a hash code of fixed length, thereby realizing feature compression and encoding. And by calculating the similarity (such as cosine similarity) of the hash codes corresponding to the semantic feature vectors of the local area of ​​the solidified pattern and the semantic feature vectors of the local area of ​​the design pattern, a quantitative evaluation of the semantic similarity between the two can be realized, thereby obtaining the hash similarity of the local areas corresponding to each group, which is used to reflect the degree of matching between the solidified pattern and the design pattern in local details. Furthermore, the hash similarity values ​​of the obtained multiple local areas are combined into a global semantic consistency representation vector, which is used to reflect the degree of semantic consistency between the solidified pattern and the design pattern in a global scope, and further evaluate the similarity between the two in global structure.

[0038] In a preferred example of the present application, the hash similarity between each group of corresponding solidified pattern local area semantic feature vectors and design pattern local area semantic feature vectors in the set of the solidified pattern local area semantic feature vectors and the set of the design pattern local area semantic feature vectors is calculated, specifically including the following steps: cascading each group of corresponding solidified pattern local area semantic feature vectors and design pattern local area semantic feature vectors to obtain a solidified pattern-design pattern local area joint semantic feature vector; performing point addition on the solidified pattern-design pattern local area joint semantic feature vector and the reciprocal of the square root of the second norm of the solidified pattern-design pattern local area joint semantic feature vector to obtain a solidified pattern-design pattern local area joint rotation offset feature vector; calculating an exponential function of the solidified pattern-design pattern local area joint rotation offset feature vector with a natural constant as the base to obtain a solidified pattern-design pattern local area joint rotation offset feature vector. A local area joint class offset prediction feature vector; point multiplying the solidified pattern-design pattern local area joint semantic feature vector with a norm and a weight hyperparameter of the solidified pattern-design pattern local area joint semantic feature vector to obtain a solidified pattern-design pattern local area joint boundary constraint feature vector; point multiplying the solidified pattern-design pattern local area joint class offset prediction feature vector with the solidified pattern-design pattern local area joint boundary constraint feature vector to obtain an optimized solidified pattern-design pattern local area joint semantic feature vector; cascading the optimized solidified pattern-design pattern local area joint semantic feature vector based on the solidified pattern local area semantic feature vector and the design pattern local area semantic feature vector to obtain an optimized solidified pattern local area semantic feature vector and an optimized design pattern local area semantic feature vector; calculating the hash similarity between the optimized solidified pattern local area semantic feature vector and the optimized design pattern local area semantic feature vector.

[0039] That is, in order to improve the calculation accuracy of the hash similarity between the local area semantic feature vector of the solidified pattern and the local area semantic feature vector of the design pattern, it is expected to improve the overall distribution regression constraint of the local area semantic feature vector of the solidified pattern and the local area semantic feature vector of the design pattern to the joint hash similarity space. Therefore, in the above preferred example, the structured norm representation of the solidified pattern-design pattern local area joint semantic feature vector which is a cascade of the solidified pattern local area semantic feature vector and the design pattern local area semantic feature vector is used as the local canonical coordinate for each eigenvalue of the solidified pattern-design pattern local area joint semantic feature vector to determine the class rotation offset of the vector of the solidified pattern-design pattern local area joint semantic feature vector relative to the eigenvalue for the class offset prediction direction of each eigenvalue of the solidified pattern-design pattern local area joint semantic feature vector as the center, and the bounding box of the vector eigenvalue distribution of the solidified pattern-design pattern local area joint semantic feature vector is used to perform eigenvalue constraints to improve the constraint of the solidified pattern-design pattern local area joint semantic feature vector under the overall regression distribution, thereby improving the calculation accuracy of the hash similarity between the solidified pattern local area semantic feature vector and the design pattern local area semantic feature vector.

[0040] In the above-mentioned maskless lithography method, the step S140 determines the detection result based on the global semantic consistency representation vector. In a specific example of the present application, the method of determining the detection result based on the global semantic consistency representation vector is to pass the global semantic consistency representation vector through a classifier-based detector to obtain the detection result, and the detection result is used to indicate whether the solidified pattern meets the design requirements. It should be understood that the detector is based on a classification algorithm, and through supervised learning training, it can learn the semantic consistency pattern between the solidified pattern and the design pattern, and realize rapid and accurate judgment of whether the solidified pattern meets the design requirements, thereby converting the global semantic consistency representation vector into a clear classification result, such as "compliant" or "non-compliant", so as to effectively improve the detection efficiency and accuracy, as well as improve the intelligence level and automation level of the production line.

[0041] In summary, the maskless lithography method according to the embodiment of the present application is explained, which uses ultraviolet light as the initial light source, exposes the image data of the microelectronic pattern to the PCB board for etching through FPGA programming and DMD, and uses machine vision technology based on deep learning to measure the global semantic consistency of the image data of the etched solidified pattern and the microelectronic pattern, so as to intelligently detect whether the solidified pattern meets the design requirements. In this way, the lithography accuracy and stability can be effectively improved, the quality of lithography products can be guaranteed, and the demand for high-precision, high-efficiency, and low-cost manufacturing in the field of microelectronics can be met.

[0042] Figure 5 FIG. 1 is a block diagram of a maskless lithography system according to an embodiment of the present application. Figure 5 As shown, the maskless lithography system 100 according to the embodiment of the present application includes: a solidified pattern digital image acquisition module 110, which is used to acquire image data of a microelectronic pattern and a digital image of a solidified pattern captured by a camera; a local area multi-scale semantic analysis module 120, which is used to perform local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern, respectively, to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern; a global semantic consistency measurement module 130, which is used to perform global semantic consistency measurement on the set of local area semantic feature vectors of the solidified pattern and the set of local area semantic feature vectors of the design pattern to obtain a global semantic consistency representation vector; and a solidified pattern detection result generation module 140, which is used to determine a detection result based on the global semantic consistency representation vector.

[0043] Here, those skilled in the art can understand that the specific operations of each module in the above-mentioned maskless lithography system have been described in detail above. Figures 1 to 4 The maskless lithography method has been described in detail, and therefore, its repeated description will be omitted.

[0044] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to being implemented by adopting the above specific details.

[0045] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the module division is only a logical function division, and there may be other division methods in actual implementation. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0046] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference to a figure in a claim should not be considered as limiting the claim to which it relates.

[0047] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0048] Finally, it should be noted that the above description has been given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A maskless lithography method, characterized in that: include: providing image data of a microelectronic pattern; Inputting image data of the microelectronic pattern into a digital micromirror device, wherein the digital micromirror device includes a plurality of rotatable mirrors; Prepare UV light source; Programming and controlling the digital micromirror device through a field programmable gate array so that a plurality of rotatable mirrors of the digital micromirror device can reflect the ultraviolet light generated by the ultraviolet light source according to instructions; Using a coated lens to modulate the ultraviolet light so that the modulated ultraviolet light irradiates the PCB board coated with a photosensitive material to obtain an exposed PCB board; Developing and etching the exposed PCB board to obtain a cured pattern; Inspecting the cured pattern based on machine vision to ensure that it meets design requirements; The curing pattern is inspected based on machine vision to ensure that it meets the design requirements, including: Acquire a digital image of the solidified pattern captured by a camera; Performing local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern respectively to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern; Performing global semantic consistency measurement on the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the design pattern to obtain a global semantic consistency representation vector; Based on the global semantic consistency representation vector, a detection result is determined.

2. The maskless lithography method according to claim 1, characterized in that: Performing local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern respectively to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern, including: Slicing the digital image of the cured pattern and the image data of the microelectronic pattern in the same image block segmentation mode to obtain a set of local area images of the cured pattern and a set of local area images of the design pattern; Performing image feature extraction on the set of local area images of the solidified pattern and the set of local area images of the designed pattern respectively to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the designed pattern; The set of local area feature maps of the solidified pattern and the set of local area feature maps of the designed pattern are respectively passed through a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the designed pattern.

3. The maskless lithography method according to claim 2, characterized in that: The image features of the set of local area images of the solidified pattern and the set of local area images of the designed pattern are respectively extracted to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the designed pattern, comprising: The set of local area images of the solidified pattern and the set of local area images of the designed pattern are respectively passed through a feature extractor based on a dilated convolutional neural network model to obtain a set of local area feature maps of the solidified pattern and a set of local area feature maps of the designed pattern.

4. The maskless lithography method according to claim 3, characterized in that: The set of local area feature maps of the solidified pattern and the set of local area feature maps of the designed pattern are respectively subjected to a multi-scale semantic feature extractor to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the designed pattern, comprising: Respectively extracting the global semantic features and the local semantic features of the local area feature map of the solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern; The global semantic feature vector of the local area of ​​the solidification pattern and the local semantic feature vector of the local area of ​​the solidification pattern are fused to obtain the semantic feature vector of the local area of ​​the solidification pattern.

5. The maskless lithography method according to claim 4, characterized in that: Respectively extracting the global semantic features and the local semantic features of the local area feature map of the solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern, including: Upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; Performing global mean pooling processing on the local area feature map of the up-sampled solidified pattern to obtain a semantic channel feature vector of the local area of ​​the up-sampled solidified pattern; Point convolution encoding is performed on the semantic channel feature vector of the local area of ​​the upsampled solidified pattern to obtain a global semantic feature vector of the local area of ​​the solidified pattern.

6. The maskless photolithography method according to claim 5, characterized in that: Extracting the global semantic features and the local semantic features of the local area feature map of the solidified pattern respectively to obtain a global semantic feature vector of the local area of ​​the solidified pattern and a local semantic feature vector of the local area of ​​the solidified pattern, further comprising: Upsampling the local area feature map of the solidified pattern to obtain an upsampled local area feature map of the solidified pattern; Performing point convolution encoding on the up-sampled solidified pattern local area feature map to obtain a channel modulated up-sampled solidified pattern local area feature map; Two-dimensional convolution coding is performed on the local area feature map of the channel-modulated up-sampled solidified pattern to obtain a local semantic feature vector of the local area of ​​the solidified pattern.

7. The maskless photolithography method according to claim 6, characterized in that: Performing a global semantic consistency measurement on the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the design pattern to obtain a global semantic consistency representation vector, including: The hash similarity between each corresponding set of the solidified pattern local area semantic feature vectors and the design pattern local area semantic feature vectors in the set of the solidified pattern local area semantic feature vectors and the set of the design pattern local area semantic feature vectors is calculated to obtain the global semantic consistency representation vector composed of multiple hash similarities.

8. The maskless photolithography method according to claim 7, characterized in that: Determining a detection result based on the global semantic consistency representation vector includes: The global semantic consistency representation vector is passed through a classifier-based detector to obtain the detection result, and the detection result is used to indicate whether the solidification pattern meets the design requirements.

9. A maskless lithography system, characterized in that: include: A curing pattern digital image acquisition module, used to acquire image data of the microelectronic pattern and a digital image of the curing pattern acquired by a camera; A local area multi-scale semantic analysis module, used for performing local area multi-scale semantic analysis on the digital image of the solidified pattern and the image data of the microelectronic pattern to obtain a set of local area semantic feature vectors of the solidified pattern and a set of local area semantic feature vectors of the design pattern; A global semantic consistency measurement module, used for performing global semantic consistency measurement on the set of semantic feature vectors of the local area of ​​the solidified pattern and the set of semantic feature vectors of the local area of ​​the designed pattern to obtain a global semantic consistency representation vector; The solidified pattern detection result generation module is used to determine the detection result based on the global semantic consistency representation vector.

Images